Fix nf in inverse matching

[giacomomagni]

Somehow the nf passed during the inverse matching is not correct:

see:

eko/src/eko/runner/parts.py

Line 152 in 50f108f

nf_match = op.nf - 1 if recipe.inverse else op.nf

[2] > /data/theorie/gmagni/N3PDF/eko_fhmv_private/src/eko/runner/parts.py(155)match()
-> nf_match = op.nf - 1 if recipe.inverse else op.nf
(Pdb++) recipe
Matching(scale=2.2801, hq=4, inverse=True)
(Pdb++) op.nf
3

This can be fixed by setting always:

 nf_match = op.nf

[alecandido]

Could you also provide a unit test?

[alecandido]

This can be fixed by setting always:

 nf_match = op.nf

Btw, at this point it should not be worth to define nf_match at all. Just use op.nf.

[alecandido]

If there is no brilliant idea for any unit test in the short term, the benchmark is definitely good enough.

We can merge.

[giacomomagni]

If there is no brilliant idea for any unit test in the short term, the benchmark is definitely good enough.

Thanks, indeed I had not a brilliant idea for a unit test, as this bug was mostly visible at PDF level.

[felixhekhorn]

If there is no brilliant idea for any unit test in the short term, the benchmark is definitely good enough.

Thanks, indeed I had not a brilliant idea for a unit test, as this bug was mostly visible at PDF level.

I do have some idea (call match with mocked ome.OME and check the outcome), but I can do in a separate PR